Saved in:
Bibliographic Details
Main Authors: So, Yerim, Kim, Jiyeong, Yoon, Jiwon, Min, Dongbo
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2605.23288
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910247585579008
author So, Yerim
Kim, Jiyeong
Yoon, Jiwon
Min, Dongbo
author_facet So, Yerim
Kim, Jiyeong
Yoon, Jiwon
Min, Dongbo
contents Recent Open-Vocabulary Action Recognition (OVAR) methods typically aggregate visual features into a global representation before computing text alignment, a process that obscures local patch information and fine-grained spatio-temporal cues. We propose Similarity Volume Aggregation (SimVA), a framework that constructs a dense 4D spatio-temporal similarity volume from patch-level visual-text similarities. SimVA constructs a spatio-temporal similarity volume over local video tokens and action classes, and employs class sampling to ensure similarity aggregation scalable to large vocabularies. The similarity volume is refined by spatial aggregation, which contextualizes local similarity patterns to improve intra-frame consistency. Motion-aware modulation further injects inter-frame variation cues, highlighting dynamically changing regions. Mamba-based temporal aggregation then models the evolution of class-conditioned similarity patterns across frames. By maintaining dense visual-text correspondence, SimVA effectively transfers CLIP to video action recognition, achieving competitive performance across zero-shot, few-shot, and base-to-novel benchmarks.
format Preprint
id arxiv_https___arxiv_org_abs_2605_23288
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Spatio-Temporal Similarity Volume Aggregation for Open-Vocabulary Action Recognition
So, Yerim
Kim, Jiyeong
Yoon, Jiwon
Min, Dongbo
Computer Vision and Pattern Recognition
Recent Open-Vocabulary Action Recognition (OVAR) methods typically aggregate visual features into a global representation before computing text alignment, a process that obscures local patch information and fine-grained spatio-temporal cues. We propose Similarity Volume Aggregation (SimVA), a framework that constructs a dense 4D spatio-temporal similarity volume from patch-level visual-text similarities. SimVA constructs a spatio-temporal similarity volume over local video tokens and action classes, and employs class sampling to ensure similarity aggregation scalable to large vocabularies. The similarity volume is refined by spatial aggregation, which contextualizes local similarity patterns to improve intra-frame consistency. Motion-aware modulation further injects inter-frame variation cues, highlighting dynamically changing regions. Mamba-based temporal aggregation then models the evolution of class-conditioned similarity patterns across frames. By maintaining dense visual-text correspondence, SimVA effectively transfers CLIP to video action recognition, achieving competitive performance across zero-shot, few-shot, and base-to-novel benchmarks.
title Spatio-Temporal Similarity Volume Aggregation for Open-Vocabulary Action Recognition
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2605.23288